This is a newly initiated research project. In traditional epidemiological studies the association between phenotype (risk factor) and disease is often biased by confounding and reverse causation. As a persons genotype is assigned by a seemingly random process, genes are potentially useful instrumental variables for adjusting for such bias. This type of adjustment combines information on the genotypeCdisease association and the genotypeCphenotype association to estimate the phenotypeCdisease association and has become known as Mendelian randomization. This project aims at developing various efficient statistical methods for Mendelian Randomiztion, ranging from study design, modeling and estimation.